In order to achieve efficient link utilization, modern radio interface protocols support Hybrid Automatic Repeat request (HARQ). A HARQ transmission is characterized by transmitting a channel-encoded (e.g., by using a Turbo Code) data unit to the receiver. The receiver replies with a feedback signal that indicates either the successful reception (ACK) or the unsuccessful reception (NACK) In case of a received NACK, the data sender retransmits the same or another so-called redundancy version of the same data unit. The HARQ receiver soft-combines all received redundancy versions for the particular data unit. For each terminal there is one HARQ entity in the transmitting unit, e.g. a base station such as an eNodeB and one HARQ entity in the receiving unit, e.g. a user equipment, UE. Each HARQ entity hosts of a number of parallel HARQ processes.
The LTE radio access is intended to use scheduled transmission, i.e., shared transmission resources can be assigned to users on a very short time scale. The scheduler deciding on the resource allocation for uplink transmission is located in the base station. The main mode of operation for the scheduler in LTE is assumed to be dynamic scheduling whereby the base station transmits scheduling messages, i.e. uplink grants or downlink assignments, to the UEs to indicate which physical resources they have been allocated for uplink transmission and downlink reception. The base station also indicates how the data transmission shall be coded and modulated in both uplink and downlink. For downlink, where an asynchronous HARQ is assumed, information comprising HARQ process id and redundancy version is included on a control channel as part of the scheduling assignment. Since for the uplink a synchronous HARQ protocol is assumed, the employed HARQ process ID and the redundancy version are coupled to the transmission timing. Thus this information is neither sent in the uplink grant nor in parallel to the uplink transmission on an uplink control channel. Instead, both sender and receiver can derive process ID and redundancy version from the transmission status.
The LTE MAC specification supports two different scheduling modes. Dynamic scheduling is the mode in which each initial (new) transmission of data is indicated via a control channel (PDCCH, Physical Downlink Control Channel) Semi-persistent scheduling (SPS) is the mode in which each initial (new) transmission is made on pre-assigned resources, i.e. the radio resources are known in the time and frequency domain.
The SPS technique thus makes assignments (downlink data transmission) or grants (uplink data transmission) for the first transmission of data superfluous and thus saves control signaling resources.
The dynamic scheduling requires relatively much signaling overhead. In order to reduce this overhead, it has been decided in 3GPP to support so called semi-persistent scheduling (SPS). In SPS the periodicity of a scheduling grant (uplink) or scheduling assignment (downlink) is configured via an RRC, Radio Resource Control, protocol. The SPS is started by transmitting a grant/assignment on the PDCCH.
In LTE the identification of UEs on the PDCCH is done by reusing the CRCr Cyclic Redundancy Check r bits. The UE identity (typically C-RNTI) is scrambled with the CRC such that when a message is transmitted on PDCCH only the UE that de-scrambles with the correct C-RNTI will have a successful CRC. Other UEs will scramble with a different C-RNTI and their CRC will not be successful. Thus they will discard the command on the PDCCH. For SPS a separate C-RNTI is used, the SPS C-RNTI. Thus the UE can distinguish if a received assignment/grant on the PDCCH is a dynamic grant/assignment or a semi persistent grant/assignment based on which C-RNTI is used. The base station is responsible for assigning the different C-RNTIs uniquely to the UEs.
Semi-persistent scheduling operation thus has the following characteristics:                Configuration of SPS is done via the RRC protocol r i.e′r the resources that might be used for new transmissions (opposed to HARQ retransmissions) are configured.        If SPS is configured r the actual activation is done with a PDCCH message. Thus r a PDCCH signal is used to activate the SPS resource usage with the RRC-configured parameters.        New HARQ transmissions are transmitted at the given SPS resource allocations.        Required retransmissions are dynamically scheduled (but still the HARQ process is a semi-persistent process. This should not be confused with a dynamic HARQ transmission where the first transmission is dynamically scheduled) i.e., a PDCCH message is used to indicate the retransmission.        
A HARQ process used for semi-persistent scheduling can still be used for dynamically scheduled transmission, i.e., a certain HARQ process is not tied to either SPS or dynamic scheduling. For example, if a dynamic HARQ process is still on-going and a PDCCH SPS activation is received (indicated by the SPS C-RNTI), the dynamic process is terminated and the SPS HARQ process overrides existing data, and since the transmission is made on pre-assigned resources, no PDCCH message is sent for the subsequent SPS transmission that follows the SPS activation. On the other hand, if an SPS process is in use and a dynamic HARQ transmission is scheduled (indicated by the C-RNTI), a PDCCH message must be sent. The PDCCH message, comprising a scheduling message, i.e. an uplink Grant or a downlink Assignment, includes a field denoted as NDI (New Data Indicator). The field size is currently 1 bit. The NDI bit is an important indicator to resolve HARQ error cases and has been originally introduced for the dynamic scheduling mode. For dynamic scheduling, i.e. where the first transmission is dynamically scheduled, the NDI bit is toggled with each new transmission. Thus the value can be either 0 or 1 for a new transmission and it will remain the same value for corresponding HARQ retransmissions for a specific MAC Protocol Data Unit, PDU. For downlink transmission this allows the UE to detect whether a new HARQ transmission starts or whether a retransmission is to be expected.
Consequently, it either flushes the HARQ buffer (in case of a new transmission) or it attempts to soft-combine the retransmission with the existing content in the soft-buffer.
For uplink transmission the NDI indicates if the UE is expected to perform a retransmission of the previously transmitted data or if it shall flush the process and obtain new data from higher layers for transmission in the dynamically assigned resource. However, it has been decided that SPS activation will use the value NDI=O and SPS retransmissions will use the value NDI=1 in the corresponding PDCCH signal. Thus, there exist two different interpretations of the NDI bit depending on the RNTI value used for addressing a User Equipment, UE, on the PDDCH. If the SPS C-RNTI is used, the NDI determines whether SPS will be activated or whether a retransmission is sent.
For dynamically scheduled HARQ processes, the NDI has a different meaning as previously explained. The decision to use the NDI bit for SPS activation and retransmission indication for SPS leads to problems, since a new interpretation is introduced.
When a HARQ process associated with SPS resources should be temporarily used by a dynamically scheduled HARQ transmission, the NDI value can not be applied as in the case where the HARQ process is exclusively scheduled dynamically.
Since the NDI bit is stored in relation to the HARQ process and the HARQ process can be either an SPS process or a dynamically scheduled process, the current specification would lead to an erroneous behavior since the NDI relating to SPS would be interpreted in the context of the dynamic HARQ process, when the first dynamically scheduled HARQ transmission takes place in a process that has been used based on SPS resources before.